Totally room-temperature solution-processing method for fabricating flexible perovskite solar cells using an Nb2O5–TiO2 electron transport layer

Flexible perovskite solar cells are new technology-based products developed by the global solar industry and are promising candidates for realizing a flexible and lightweight energy supply system for wearable and portable electronic devices. A critical issue for flexible perovskite solar cells is to achieve high power conversion efficiency (PCE) while using low-temperature solution-based technology for the fabrication of a compact charge collection layer. Herein, we innovatively introduce niobium ethoxide as a precursor additive to TiO2 NCs, which allows realization of an Nb2O5–TiO2 electron transport layer (ETL). The presence of Nb2O5 remarkably enhances electron mobility and electrical conductivity of the ETLs. In addition, uniform perovskite films are prepared by an annealing-free solution-based method. The excellent performance of the cell is attributed to its smooth film surface and high electron mobility, and performance is verified by the effective suppressions of charge recombination and time-resolved photoluminescence. PCEs of 15.25% and 13.60% were obtained for rigid substrates (glass/fluorine-doped tin oxide) and an indium tin oxide/PET (poly(ethylene terephthalate)) flexible substrate by using a totally room-temperature solution-processing method, respectively.


Introduction
Perovskite solar cells have developed very rapidly over the last several years and could be in considerable demand for various applications in the future photovoltaic market. [1][2][3][4][5][6] Numerous scientic research professionals have accelerated increases in power conversion efficiency (PCE) of perovskite solar cells, and at present, the certied world record for a PCE is 22.1%. 7 As a light absorption layer, APbX 3 (A ¼ CH 3 NH 3 , (NH 2 ) 2 CH 2 or Cs, and X ¼ I, Br, or Cl) is considered as the most promising replacement for silicon solar cells; APbX 3 has outstanding properties such as strong light absorption, weakly bound excitons, long-range charge-carrier diffusion, and apparent tolerance to defects. [8][9][10] A state-of-the-art photovoltaic solar cell (PSC) is prepared in a conventional n-i-p device conguration consisting of an n-type oxide semiconductor capped with a perovskite absorber and a hole-transport layer (HTL). The electron transfer layer (ETL) of PSCs plays important roles in extracting electrons and blocking holes from perovskite. Metal oxide materials such as TiO 2 , ZnO, and SnO 2 are widely used in most studies aimed at achieving high efficiency because of their environmentally friendly nature, wide band gap, high electron mobility, and good stability. [11][12][13][14] A large number of PSCs based on TiO 2 require a high-temperature sintering process for their crystallization or removal of the dispersion medium. [8][9][10]15 However, TiO 2 , ZnO, and SnO 2 ETLs obtained using high-temperature treatment cannot be applied to exible perovskite solar cells. Recently, methods have been developed for fabricating low-temperature processable TiO 2 ETLs for planar PSCs, such as atomic layer deposition 16 and magnetron sputtering. 17 Compared to vacuum methods, solution-based techniques are generally more cost-effective and scalable and can be used to achieve roll-to-roll process-ability. Therefore, it is imperative to develop effective methods that can restriction migration, improve electron mobility of an ETL, and reduce trap-state density within the perovskite material, thereby eliminating hysteresis and improving the efficiency of PSCs. However, perovskite materials such as CH 3 NH 3 PbI 3 and CH(NH 2 ) 2 PbI 3 are mostly used for PSCs, and these materials practically need annealing for 10-60 min to form the black crystalline photoactive layer. [18][19][20][21][22][23] Annealing process is not convenient for mass production because it requires additional equipment and increases energy consumption. One method to address this problem is to employ annealing-free processing, 15 which can save energy and facilitate industrial production. Herein, we propose a facile strategy to prepare a highly efficient perovskite solar cell via room-temperature solution processing. TiO 2 lms are fabricated by spin coating a colloidal solution of anatase TiO 2 nanocrystallines (NCs) prepared via a low-temperature sol-gel method on a substrate. Considering improved performance of the lm, we employed niobium ethoxide as a precursor additive. Nb 2 O 5 is considered a better ETL for a perovskite solar cell due to its higher carrier mobility and conduction band edge position. Here, niobium ethoxide is introduced into the TiO 2 dispersion as an aid dispersant. Niobium ethoxide facilitates spontaneous coalescence of the TiO 2 NCs, thereby forming a stable dispersion. The stable dispersion is spin-coated onto the conductive base lm without annealing; only ultraviolet (UV) treatment for 15 min yields a uniform and dense Nb 2 O 5 -TiO 2 layer. It is expected that the Nb 2 O 5 formed in situ would passivate the grain boundary of the TiO 2 NCs, and further form a dense and uniform lm. The exceptional performance of the layer is attributed to the excellent optical and electronic properties of the Nb 2 O 5 -TiO 2 material, such as a smooth surface and high electron mobility; these properties make the material a better growth platform for a high-quality perovskite absorber layer. In addition, MAPbI 3 lms are deposited as a light absorption layer via one-step spin-coating by a simple annealing-free process.

Synthesis of TiO 2 NCs and preparation of TiO 2 dispersion with niobium ethoxide
TiO 2 NCs were synthesized by a non-hydrolytic sol-gel reaction according to a modied procedure. 24 The resulting precipitates were washed by adding excess ethanol and diethyl ether and puried by centrifugation at 3000 rpm for 5 min. This washing procedure was repeated thrice. To obtain the TiO 2 colloidal solution ($5 mg mL À1 ), the washed TiO 2 NCs were dispersed into anhydrous methanol, and ultrasonic treatment was carried out for several hours. To obtain a niobium ethoxide/TiO 2 mixed precursor suspension for spin coating, the puried TiO 2 NCs (5 mg mL À1 ) were re-dispersed in ethanol at the desired niobium ethoxide concentration. Nb 2 O 5 -TiO 2 ETLs were fabricated by spin-coating at 3000 rpm under ambient conditions, and the lms were free from annealing. The samples were treated again with UV-ozone for 15 min before perovskite deposition.

Device fabrication
Pre-patterned transparent conducting oxide substrates were sequentially cleaned using ethanol, acetone, isopropanol, and ethanol separately in an ultrasonic bath for 20 min each and then dried under owing nitrogen. Fluorine-doped tin oxide (FTO) substrates underwent UV-ozone treatment (Model UV-03 UVO 3 cleaner) for 15 min before they were used for spin-coating ETLs.
A highly dispersed solution of TiO 2 NCs and TiO 2 NCs with niobium ethoxide in ethanol were dropped onto substrates and immediately spin-coated at a speed of 3000 rpm for 30 s. The samples again underwent UV-ozone treatment for 15 min before perovskite deposition. Then, the TiO 2 (Nb 2 O 5 -TiO 2 )coated substrates were transferred immediately to a nitrogen-lled glovebox for the deposition of perovskite lms.
A MAPbI 3 solution was prepared according to an annealingfree process reported by Fang et al. 15 The MAPbI 3 precursor solution (1.2 M) was prepared using a mixed solvent of DMAc and NMP in a volume ratio of 5 : 1. Perovskite lms were deposited onto the TiO 2 or Nb 2 O 5 -TiO 2 substrates according to a two-step spin-coating procedure. In the rst step, spin coating was carried out at 1000 rpm for 20 s with an acceleration of 200 rpm s À1 . In the second step, spin coating was carried out at 5000 rpm for 45 s with an acceleration of 1000 rpm s À1 . During the second step, chlorobenzene was dropped onto the spinning substrate at 35 s before the end of the procedure. At the end of the second step, a dark perovskite lm was directly formed. A spiro-OMeTAD HTL was prepared according to a process reported by Yang et al. 25 The HTL was fabricated as follows: a spiro-OMeTAD solution (90 mg mL À1 ) was dissolved in chlorobenzene using 36 mL 4-tert-butylpyridine and 22 mL lithium bis(triuoromethylsulfonyl)imide (520 mg mL À1 ) as the dopants in acetonitrile. The spiro-OMeTAD solution was spincoated onto the perovskite lms at 3000 rpm for 30 s. Finally, an 80 nm thick gold coating was deposited using a thermal evaporator.

Characterization
Morphologies of the TiO 2 lms were characterized by eldemission scanning electron microscopy (FESEM, Zeiss Supra 55). The z-potential of the TiO 2 NCs with different concentrations of niobium ethoxide was characterized by using a size analyzer (Zetasizer Nano ZS ZEN3600 instrument, Malvern Instruments) at room temperature with a 633 nm laser. The UVvis transilluminator spectra of the samples were recorded on a spectrophotometer (UV5800). High-resolution transmission electron microscopy (HR-TEM) was carried out using an electron microscope (JEM-2100, JEOL Ltd., Japan). The root-meansquare (RMS) roughness and topography images of the lms were obtained via atomic force microscopy (AFM, Veeco Dimension V). The quality and crystalline structure of the samples were conrmed by q-2q X-ray diffraction (XRD) using an X-ray diffractometer (D/max 2500 PC) with a Cu Ka radiation source. A photoluminescence (PL) system (DeltaFlex, Horiba Ltd.) was used to measure the time-resolved PL (TRPL) decay. Photovoltaic performance of the solar cells was measured using a multisource meter (Model 2400, Keithley, Cleveland, OH, USA) under one sun (AM 1.5G, 100 mW cm À2 ) illumination, which was achieved by using a solar simulator (500 W Xe lamp) (XES-40S1, San-Ei Electric Co., Ltd., Japan) as the light source. The device area of 0.07 cm 2 was dened by a metal mask. All devices were scanned with a reverse and forward under standard test procedure at a scan rate of 0.2 V s À1 . X-Ray photoelectron spectroscopy (XPS) was performed on a photoelectron spectrometer (ESCALAB 250Xi, Thermo Fisher Scientic).

Characterization of TiO 2 and Nb 2 O 5 -TiO 2 lms
In the spin-coating process, a stably distributed precursor solution of TiO 2 NCs in solution is a prerequisite to form uniform and void-free TiO 2 thin lms. The niobium ethoxidecapped TiO 2 NCs have a uniform dispersion in ethanol, are stable for months, and have better anti-settleability properties (Fig. S1 †). The z-potential is an important factor reecting colloid stability. The z-potential was measured, and the results are shown in Fig. 1. The gure indicates that the z-potential value for TiO 2 is 16.3 mV. The z-potential increases with the addition of niobium ethoxide. It demonstrated that the niobium ethoxide plays an important role in the dispersion of TiO 2 NCs. Powder XRD measurements of a TiO 2 sample revealed typical diffraction peaks of anatase TiO 2 (Fig. 2a) To gain further insights into the nanoscale morphology and nanocrystal structures, the morphology of as-synthesized TiO 2 NCs was investigated using the HR-TEM images. The HR-TEM images (Fig. S2a †) revealed that the TiO 2 NCs are around 5-10 nm in diameter, which is consistent with the value obtained by Scherrer peak width analysis. In addition, the selected area electron diffraction pattern for the TiO 2 NCs (Fig. S2b †) conrms the high crystallinity of the TiO 2 NCs. Fig. 2b shows the optical transmission spectra of TiO 2 and Nb 2 O 5 -TiO 2 . Both the materials show excellent transmittance in the wavelength range of 400-800 nm. The resultant ETLs coated on the FTO glass substrates shows good optical transparency and higher transmittance than the bare FTO glass, with the transmittance being greater than 80% in the entire visible region. The high transparency of the Nb 2 O 5 -TiO 2 ETLs is very conducive to the light absorption of the perovskite layer and improves light harvesting. No apparent difference was found between the transmittances of different Nb 2 O 5 -TiO 2 lms deposited on the FTO substrates.
The composition and bonding type of the Nb 2 O 5 -TiO 2 lm were measured using XPS. A typical XPS spectra of Nb 2 O 5 -TiO 2 is shown in Fig. 3a. Clearly, the O, Ti, and Nb peaks are located at $530.59, $458.99, and $207.51 eV, respectively. The highresolution Ti 2p (Fig. 3b) spectrum reveals two different peaks located at 459.04 and 464.69 eV, which correspond to Ti 2p 3/2 and Ti 2p 1/2 , respectively; accordingly, a spin-orbit coupling of 5.65 eV is obtained, which is the signature of Ti 4+ . As shown in the Nb 3d core level spectra (Fig. 3c), Nb 3d 5/2 and Nb 3d 3/2 peaks are located at 207.54 and 210.20 eV, respectively, indicating the presence of ve-valent niobium in the deposited lms. 27 The main binding energy of 530.3 eV is attributed to O 1s (Fig. 3d) which indicates the O 2À state in TiO 2 and the peak at the higher binding energy of 531.4 eV is attributed to surface oxygen-group absorbance or hydroxyl groups. 28-31 Fig. 4a and b show the top-view SEM images of the TiO 2 and Nb 2 O 5 -TiO 2 lms. The Nb 2 O 5 -TiO 2 lm is uniform and dense and shows no apparent pinholes, indicating the high quality of the lm. The surface morphology of the Nb 2 O 5 -TiO 2 lm is not like to that of the pristine TiO 2 lm. The results of simultaneous energy-dispersive spectroscopy (EDS) of the Nb 2 O 5 -TiO 2 lm are shown in Fig. S3. † When excess niobium ethoxide (30% and 40%) is added, the surface morphology indicates the formation of a porous surface with a large number of pinholes, directly causing a deterioration in the quality of the thin lm (Fig. S4 †). Fig. 4c and d shows the AFM height images of the TiO 2 and Nb 2 O 5 -TiO 2 lms. The RMS roughness decreased from 13.8 to 10.5 nm because of the introduction of Nb 2 O 5 , indicating that the Nb 2 O 5 -TiO 2 lms have at surfaces. A smooth surface is essential for growing high-quality perovskite lms, reducing surface defect trap interfaces with ETLs and HTLs, and enhancing charge extraction at the interface between the ETL and the perovskite layer. [32][33][34] To investigate the effect of introducing Nb 2 O 5 on the electrical properties, the electron mobility was studied by the space charge limited current (SCLC) method using an electron-only device structure. The sample structure used for this measurement was FTO/PCBM/TiO 2 (15% Nb 2 O 5 -TiO 2 )/PCBM/Ag. The details are shown in the ESI. † Fig. S5 † shows the current density-voltage (J-V) curves for the TiO 2 and Nb 2 O 5 -TiO 2 lms tted using the Mott-Gurney law. 34,35 It's apparent that the electron mobility of the ETL lm increases considerably because of the introduction of Nb 2 O 5 . The electron mobility increases from 7.09 Â 10 À4 to 1.14 Â 10 À3 cm 2 V À1 s À1 , which is very close to previous reports that TiO 2 modied cra with annealing process. 36

Photovoltaic performance
The above measurements show that the quality of the TiO 2 NCs lm improves considerably because of the introduction of Nb 2 O 5 ; e.g., the lm has a smoother surface and enhanced electron mobility. Next, PSCs were designed and fabricated using the TiO 2 and Nb 2 O 5 -TiO 2 ETLs. Fig. S6 † shows the detailed device structure, in which FTO is employed as the anode, the TiO 2 or Nb 2 O 5 -TiO 2 lm as the ETL, MAPbI 3 as the absorber layer, spiro-OMeTAD as the HTL, and a gold layer as the cathode. A cross-sectional SEM image of the completed device architecture is shown in Fig. S7. † The smooth morphology of the Nb 2 O 5 -TiO 2 lms is benecial for forming  highly crystalline and compact perovskite lms. Fig. S8 † shows the top-view SEM images of the perovskite lm, which exhibits smooth surfaces, big crystalline size, and good coverage.  Fig. 5a shows the J-V curves for the champion devices based on both TiO 2 and Nb 2 O 5 -TiO 2 ETLs measured in the reverse and forward scan directions. Compared to the control device, the photovoltaic performance of these champion devices is considerably better; the larger J sc and FF are attributed to the improved electron mobility and better hole blocking effect of the Nb 2 O 5 -TiO 2 ETL, and the high V oc may be due to the reduced charge recombination and improved electron extraction. [37][38][39] Fig. 5b shows the incident photon-tocurrent efficiency spectra for various ETLs. The integrated current density value for the pristine TiO 2 -based cell is 18.92 mA cm À2 , and it increases to 19.50 mA cm À2 for the Nb 2 O 5 -TiO 2based device; that value is in good agreement with the J-V measurement value. Performance statistics for 30 individual cells with TiO 2 and Nb 2 O 5 -TiO 2 ETLs are shown in Fig. 6. Clearly, PCEs show a narrower distribution with a smaller standard deviation for the Nb 2 O 5 -TiO 2 -based cells, indicating good reproducibility.

Recombination
To gain insight into the electron extraction and transport mechanism, the steady-state PL and TRPL were measured for the perovskite absorber layer deposited on both the ETL-based substrates. Fig. 7a shows that the spectral peak for the FTO/ perovskite sample at 766 nm apparently has the highest PL intensity, indicating serious recombination occurring in the sample. The FTO/Nb 2 O 5 -TiO 2 /perovskite sample has the lowest PL intensity, even lower than that of the FTO/TiO 2 /perovskite sample. Interestingly, compared to the PL peak for the control samples, the peak for the FTO/Nb 2 O 5 -TiO 2 /perovskite lm exhibits a considerable blue-shi. 40 This suggests that Nb 2 O 5 -TiO 2 substrate is favorable to obtain higher-quality perovskite lms than the control group. Perovskite with higher crystallinity and considerably less trap density than the corresponding bandgap will decrease (i.e., the conduction band minimum will move down). 41 Fig.  7b shows the time-resolved

Flexible perovskite solar cells
The totally room-temperature UV process for efficient ETLs is very suitable for fabricating high-performance exible PSCs. The excellent performance of uniform and dense Nb 2 O 5 -TiO 2 ETLs prompted us to fabricate exible perovskite solar cells using exible substrates. Nb 2 O 5 -TiO 2 ETLs were successfully fabricated on indium tin oxide (ITO)/PET substrates, and perovskite, spiro-OMeTAD, and a gold electrode were sequentially deposited using the same methods as those used for the FTO glass-based devices. Fig. 8a shows the J-V curves for the perovskite solar cells fabricated using exible ITO/PET substrates, and the inset illustrates a photograph of a exible perovskite solar cell fabricated using exible ITO/PET substrates. For the exible device, J sc is 20.04 mA cm À2 , V oc is 0.99 V, and the FF is 0.69, giving the best PCE of 13.60%. The PCE is lower than that of the rigid device because of the decreased J sc , V oc and FF; the decrease is probably caused by the higher series resistance and lower transmittance of the ITO/PET substrate in the short wavelength spectrum. 17 Fig. 8b shows the J-V curves for the exible device aer it is recovered from the given bending radius. The key J-V parameters of the devices are summarized in Table S3. † Aer the device is bent with R ¼ 10, 5,  and 3 mm, the PCE values degenerate to 13.03%, 10.70%, and 3.39%, respectively. As shown in Fig. 8b, the performance of the exible device does not show serious degradation when the bending radius is 10 mm. When the bending radius is 3 mm, the brittle ITO breaks 44 and the PCE is greatly reduced simultaneously, indicating that the exible devices show good mechanical stability and the Nb 2 O 5 -TiO 2 ETLs is a promising electron transport material.

Conclusion
We have demonstrated that Nb 2 O 5 -TiO 2 is an excellent ETL material for perovskite solar cells, with the champion cell showing a considerably higher PCE (15.25%) than that of devices based on a pristine TiO 2 ETL and that of a rigid substrate (13.47%). In the proposed process, niobium ethoxide facilitates the spontaneous coalescence of the TiO 2 NCs, thereby forming Nb 2 O 5 -TiO 2 ETL. Our results suggest that lowtemperature solution-processed Nb 2 O 5 -TiO 2 could be a good ETL candidate for producing efficient perovskite solar cells. Our facile strategy is highly suitable for fabricating high-performance exible PSCs because it does not need a hightemperature process and can easily modify the ETL via the direct addition of a reagent. This approach will pave the way for further advances in exible PSCs and is feasible for large scale roll-to-roll processing.

Conflicts of interest
There are no conicts to declare.